Monitoring Turfgrass Seasonality with Bayesian Hierarchical Models
HKU CompSci 2024 Final Year Project #24005
Chan Yan Tak (3035927635)
Nip Hok Leung (3035957240)
⠀
Supervised by Dr. Choi Yi King
More about turfgrass
Turfgrasses are narrow-leaved grass species that can uniformly cover the ground, creating a dense, green lawn or turf. They are widely used in residential, commercial, and recreational areas, offering both aesthetic appeal and functional benefits such as soil erosion prevention, heat island mitigation, carbon sequestration, pollutant absorption, and noise reduction. Like all plants, turfgrass has growth cycles influenced by seasonal environmental changes. Understanding this seasonality is crucial both for turf breeding and management: For turf breeders, it informs selection of parental plants and guides breeding objectives. For turf managers, it allows maintenance practices to be optimized, and ensures the turfgrass fulfills aesthetic and functional requirements.
*Problems
During cultivar evaluations, trained raters rate the quality of turfgrasses on a scale from 1 to 9. However…
- Rating is subjective among raters, and may not be accurate
- Large patches of nearby plots can perform significantly better or worse than expected, but was not accounted for during rating, leading to bias results.
*Solution
- Incoperate Machine Learning in
- Case studies that celebrate architecture.
- Exclusive access to design insights.
Upcoming schedule:
| Date: | To do: |
| September 2024 | – Define the scope for the project and set project plan – Set up web page for project – Reading: Gain familiarity with previous work, A latent scale model to minimize subjectivity in the analysis of visual rating data for the National Turfgrass Evaluation Program – Reading: Gain familiarity with Gaussian Processes, starting at A Visual Exploration of Gaussian Processes |
| October 2024 | – Reading: Gain familiarity with Item Response Theory models for modeling rater behavior, starting at (PDF) A Rating Scale Formulation for Ordered Response Categories – Learn to use PyMc through reading documentation, trial and error, start to work on model implementation – Set up working environment for project |
| November 2024 | – Bayesian Hierarchical Model implemented and sanity checked – Start to experiment with and develop visualizations of results |
| December 2024 | – Continue to develop result visualizations – Variations of model tried and compared using ELPD-LOO – Prepare interim report/presentation |
| January 2025 | – Reading: Gain familiarity with the topic of Turfgrass growth potential prediction using temperature, starting at Using temperature to predict turfgrass growth potential (GP) and to estimate turfgrass nitrogen use – Model validation using parameter recovery on synthetic dataset – Collect data for rating scale alignment across different trial locations |
| February 2025 | – Model growth potential curve in Python – Development of proof of concept for rating scale alignment |
| March 2025 | – Organize code into easy to use Python package – Try model on other datasets – Complete any delayed tasks from previous months due to unforeseen challenges |
| April 2025 | – Code cleanup – Prepare final report and presentation |